Vertical Compute is a Belgian deeptech semiconductor startup developing a new class of 3D memory architecture in which vertical memory elements are integrated directly above compute logic within a single 300mm wafer manufacturing process. By shrinking the distance data must travel from centimeters between chips to nanometers within a single die stack, the company aims to address one of the most stubborn bottlenecks in AI hardware, the so-called memory wall.
The company spun out of Belgian research institute imec in 2024 and is based in Louvain-la-Neuve, with a roughly 25-person team across Belgium and France. Its technology is designed to be delivered as chiplets that integrate seamlessly into existing CPU, GPU and custom AI accelerator architectures, acting as an enabling memory layer rather than a wholesale replacement for compute cores.
Vertical Compute's seed round totals €57M, made up of a €37M extension announced in 2026 on top of an earlier €20M tranche. Quantonation led the latest round, with participation from Flanders Future Techfund (managed by PMV), Wallonie Entreprendre, Sambrinvest, Noshaq, InvestBW, Drysdale Ventures and Kima Ventures. Existing investors Eurazeo, XAnge, Vector Gestion, imec.xpand and imec also reinvested, an unusually deep cap table for a hardware company at this stage.
In just over a year the team has taped out a first test chip that integrates vertical memory stacks directly on compute logic within a single wafer manufacturing flow, a meaningful technical milestone for any 3D memory startup. The new capital is earmarked for the transition from chip validation to commercial chiplet deployment with early customers building next-generation AI accelerators.
Vertical Compute operates in a high-stakes segment alongside HBM vendors, near-memory compute startups and other 3D integration efforts inside large foundries. Its differentiation rests on the monolithic, single-process integration approach, deep imec relationships and a chiplet packaging strategy that fits with the broader chiplet ecosystem rather than asking customers to redesign their architectures.